Insurance / Insurtech · Deep dive

Kalepa

An AI underwriting workbench that reads messy commercial-insurance submissions — broker emails, SOVs, loss runs, supplemental apps — and surfaces the hidden exposures underwriters miss, betting that a human-in-the-loop 'Copilot' that keeps the underwriter deciding can out-compete both the autonomous-agent insurtechs raising four times its capital and the carriers building ingestion in-house.

emerging

The question that decides it: Kalepa's wedge is a human-in-the-loop underwriting Copilot — it reads the messy submission, cross-references third-party data, scores the risk, and hands the underwriter a decision with the reasoning shown, keeping a person in the loop rather than replacing them. Does a decision-support copilot stay defensible as (a) rivals like Sixfold and Federato push toward autonomous 'AI Underwriter' agents that aim to remove the underwriter entirely, (b) policy-admin incumbents (Applied Systems, which bought both Planck and Cytora in 2024-2025; Guidewire, which now backs Sixfold) bundle submission ingestion into systems carriers already run, and (c) large carriers' own data and off-the-shelf LLMs let them build the ingestion layer in-house — or does Kalepa's out-of-the-box accuracy on genuinely unstructured specialty risk, proven on $10B+ carrier books, become the thing everyone else has to match?

My take ★★ Very interesting ★ Interesting − Maybe ✗ Not interesting

Snapshot

Kalepa sells Copilot, an AI underwriting workbench for commercial property-and-casualty insurers that reads the unstructured chaos of a submission — broker emails, statements of value, loss runs, ACORD forms, supplemental applications — digitizes it, cross-references first- and third-party data, flags the hazards and exposures a human might miss, and hands the underwriter a decision with the reasoning shown. Founded in 2018 in New York by two Applied Predictive Technologies alumni with no insurance background, it has raised only about $16 million — a $14 million Series A led by Inspired Capital in September 2021 on top of a small IA Ventures seed — yet counts $10-billion-plus carriers including Arch, Munich Re and affiliates of W. R. Berkley among its customers, alongside hypergrowth MGAs like Bowhead Specialty and small regional mutuals. It matters now because commercial underwriting is the last large insurance workflow still done largely by hand on bad data, and because Kalepa is trying to win it as a capital-light copilot at the exact moment rivals raising four-to-five times its money are betting on fully-autonomous AI underwriters instead.

Founding story

Kalepa is a case of outsiders diagnosing an industry from first principles. Paul Monasterio trained as a physicist — nuclear engineering at Berkeley, a computational-physics PhD at MIT — before spending years at Applied Predictive Technologies, the causal-analytics company that helped large enterprises test decisions before making them, and then at Facebook running measurement for monetization. Daniel Hillman came from military intelligence in the IDF, where decisions were high-consequence and made under uncertainty, then Penn engineering, Mastercard and Monitor Deloitte, overlapping with Monasterio at APT. The two kept circling the same observation: enormous, expensive business decisions are routinely made on incomplete, unstructured information, and nobody has built the tooling to fix it.

Commercial insurance underwriting was the sharpest version of that problem. An underwriter deciding whether to bind a policy — and at what price — is handed a pile of inconsistent documents from brokers, must reconcile them, hunt for hidden exposures (a demolition contractor moonlighting as a roofer; a restaurant with a prior fire; a fleet with a bad loss history), and price accordingly, often in a time-pressured market where the fastest credible quote wins. Neither founder came from insurance, which they treat as the point: they saw underwriting not as a workflow to digitize incrementally but as a decision-quality problem — the exact thing APT and intelligence work had taught them to attack. They launched Kalepa in 2018, raised a modest IA Ventures seed, and spent the early years earning the right to sit inside real carriers’ underwriting desks.

How it works

When a submission lands, Copilot goes to work before the underwriter does. First it digitizes the mess: proprietary models parse broker emails, SOVs, loss runs, applications and supplemental forms into structured data, regardless of format. Then it enriches and cross-references — pulling third-party signals like news, business reviews, geographic and catastrophe risk, government and regulatory databases, and the carrier’s own other vendor feeds — to build a picture of the actual business behind the application, not just what the broker wrote down. It then prioritizes: the AI engine surfaces the submissions most likely to bind and fit appetite, and filters out those that violate the insurer’s guidelines, so underwriters spend time on the accounts worth winning.

The design philosophy is deliberately human-in-the-loop. Copilot does the reading, the cross-referencing and the scoring; the underwriter does the deciding — and, crucially, can see why the machine flagged what it flagged, rather than being handed an opaque score. It is pitched as working out-of-the-box, so a carrier reaps value from day one rather than after a multi-quarter data-integration project. One uncomfortable detail from employee reviews is worth stating plainly: at least some of the enrichment has historically leaned on a human review layer behind the AI — a reminder that “reads every submission accurately” is genuinely hard, and that the line between automation and augmented human labor is where a lot of the operational cost, and the competitive question, actually sits.

Product and business overview

Copilot is sold as an underwriting workbench with several named capabilities: automated intake (turning unstructured submissions into structured data), intelligent triage (prioritizing and declining against guidelines), risk evaluation (surfacing exposures from first- and third-party data), decision support (a streamlined interface where the underwriter reviews flags and corresponds with brokers in one place), portfolio monitoring, and workflow automation. The moat the company leans on is breadth of line coverage combined with accuracy on genuinely messy risk: Copilot supports excess and surplus (E&S) lines, primary casualty, property, commercial auto and fleet, and in a 2024-2025 expansion added management liability — the specialty, hard-to-standardize corners of the market where unstructured judgment matters most and where generic automation struggles.

Business model and pricing

Kalepa is enterprise SaaS. Carriers and MGAs pay subscription fees for access to Copilot, with the company describing pricing as subscription-based and potentially performance-linked; it does not publish a price list, and deals are quoted per carrier. The go-to-market spans the full size range — from $10B+ carriers like Arch and Munich Re, to hypergrowth MGAs like Bowhead Specialty and Paragon, to small regional insurers like North Star Mutual — which suggests a land-and-expand motion: prove Copilot on one line or one operating company (as with Admiral, a Berkley company, before the broader W. R. Berkley expansion), then widen across lines and affiliates. The out-of-the-box positioning is the commercial wedge: because Copilot is pitched as delivering value without a long integration, Kalepa can sell on fast, demonstrable ROI rather than on a platform migration.

Traction over time

The shape of the story is unusual: modest capital, marquee logos. Kalepa has not announced a Series B in the roughly five years since its Series A, and third-party estimates put ARR around $5M for 2024 and headcount under 60 in 2026 — small numbers. But the customer list (Arch, Munich Re, W. R. Berkley affiliates) is disproportionately blue-chip for a company that has raised $16M, and the InsurTech100 recognition in three separate years signals durable relevance. The bull read is extreme capital efficiency; the bear read is that five years without a priced up-round, on ~$5M ARR, is a long time to stay small in a category where rivals are raising aggressively.

Market analysis

The addressable market is large and being reshaped by AI in real time. Mordor Intelligence pegged the underwriting-software market at roughly $7.15B in 2025, growing about 12.5% annually to ~$12.9B by 2030; the broader commercial-insurance-software market is put around $11B in 2025 rising toward ~$29B by 2034 (multiple analysts, 2025). The structural forces all point the same way: a wave of underwriter retirements is draining institutional risk-judgment from carriers; submission volumes are rising faster than headcount; combined-ratio pressure makes risk selection and speed-to-quote existential; and foundation models have, for the first time, made unstructured-document understanding cheap enough to attack the intake bottleneck. The countervailing force is that this same cheapness lowers the barrier for everyone — carriers, core-system vendors and a crowd of insurtechs — which is why the category has gone from niche to contested in about three years.

Competitive intel

The competitive set splits into two camps, and Kalepa sits awkwardly between them (full profiles in the sidebar). On one side are the autonomous-agent challengers — Sixfold ($50M+ raised, building an end-to-end “AI Underwriter,” now backed by Guidewire) and Federato ($80M raised, RiskOps portfolio orchestration) — both better funded than Kalepa and both selling a more expansive vision: remove or reorient the underwriter, don’t just assist them. On the other side are the data-and-scoring specialists and the incumbents absorbing them: Gradient AI ($87M raised, loss-prediction scoring), and most tellingly Planck ($71M raised, acquired by Applied Systems for $300M in July 2024) and Cytora ($41.5M raised, acquired by Applied Systems in September 2025). Those back-to-back acquisitions are the sharpest competitive signal in the whole set: the submission-intake and enrichment layer Kalepa lives in is being rolled up into policy-admin platforms carriers already own. Kalepa’s differentiation — genuine depth on messy specialty submissions, proven out-of-the-box on $10B+ carrier books — is real, but it is being pressed from above by autonomous agents, from the side by better-capitalized scorers, and from below by incumbents bundling ingestion for free.

History and evolution

• 2018 — Founded in New York by Paul Monasterio and Daniel Hillman, two APT alumni approaching underwriting as a decision-quality problem.
• ~2019-2020 — Small seed led by IA Ventures; early carrier pilots.
• Sep 2021 — $14M Series A led by Inspired Capital, with IA Ventures and fintech-operator angels (Gokul Rajaram, Jackie Reses, Henry Ward); total raised reaches ~$16M.
• 2022-2023 — Named to InsurTech100; expands Copilot line coverage across E&S, casualty, property and commercial auto.
• 2024 — getLatka estimates ~$5M ARR; adds fleet and management-liability support; competitors’ consolidation begins (Applied buys Planck).
• 2024-2025 — Deploys at Admiral (a Berkley company); Applied buys Cytora; category consolidation accelerates.
• 2025-2026 — W. R. Berkley affiliate expands Copilot across operating companies; Bowhead Specialty, Paragon and MGA Beyond Risk deploy; back on the 2025 InsurTech100. No Series B announced; headcount ~58 (May 2026).

The notable “stumble” is negative space: no disclosed priced financing since 2021 and a sub-$10M revenue estimate five years in. Whether that reflects discipline or difficulty is the central ambiguity of the company.

What people say

The case for. On Glassdoor, Kalepa carries roughly a 4.3/5 rating with about 83% of reviewers recommending it and compensation rated highly (~4.5/5). The recurring praise is a genuine entrepreneurial culture: smart people, real ownership, trust to run your own book, and fast career progression for those who perform. Customers, in the company’s own case studies and trade coverage, describe Copilot as a “must-have” workbench that raises underwriting quality and consistency, and the blue-chip logo list — Arch, Munich Re, W. R. Berkley affiliates, Bowhead — is the strongest third-party validation available: large, sophisticated carriers do not deploy underwriting tooling casually. Three InsurTech100 selections reinforce that this is a taken-seriously vendor, not a demo-stage startup.

The complaints. The negatives are specific and pointed. The most damaging employee critique is that the “AI” is partly a marketing veneer — that beneath the interface sits a series of third-party tools and in-house patches propped up by a crowdsourced team of human reviewers, with the tech stack described as underwhelming and the company as too cost-conscious to buy enough software licenses. Reviewers also flag the classic startup grind: long hours, high performance expectations, little supervision, and a culture “not for those looking to do minimal work.” The strategic critique compounds the operational one: on ~$16M raised and ~$5M estimated ARR, Kalepa is small and under-capitalized relative to Sixfold, Federato and Gradient AI, and the layer it competes in is actively being consolidated by Applied Systems. A copilot that leans on human reviewers behind the scenes is exactly the thing autonomous-agent rivals claim to obsolete.

Outlook: the open question

Kalepa is a genuinely intriguing anomaly: a capital-light insurtech that has won trust from some of the most sophisticated carriers in commercial P&C while raising a fraction of what its rivals have. The product clearly works well enough for Arch, Munich Re and W. R. Berkley affiliates to deploy and expand it — that is not nothing, and it is more than most better-funded competitors can claim. But the company sits at the intersection of three converging pressures, and its size gives it little margin for a wrong turn.

Kalepa wins if the human-in-the-loop copilot turns out to be the durable form factor — if underwriters and their regulators keep wanting an assistant that shows its reasoning and leaves a person accountable rather than a black-box autonomous underwriter, if Copilot’s out-of-the-box accuracy on messy specialty lines stays a step ahead of what carriers can build in-house or buy bundled from Applied/Guidewire, and if its blue-chip deployments expand across lines and affiliates into real eight-figure ARR that finally justifies (or removes the need for) a priced up-round. It struggles if the autonomous-agent thesis wins — if Sixfold’s “AI Underwriter” and Federato’s RiskOps prove that carriers will trust agents to underwrite with less human oversight, if the submission-intake layer Kalepa occupies keeps getting commoditized into the core systems carriers already run, or if five-plus years without fresh priced capital on ~$5M ARR simply means the company gets out-invested and out-distributed before efficiency turns into scale. The tell to watch over the next 12-18 months is straightforward: a Series B (or a credible reason it doesn’t need one), and whether the W. R. Berkley relationship expands from one operating company into many. If capital-efficient depth compounds into carrier-wide, multi-line deployments, Kalepa’s restraint looks like the smartest bet in the category. If it stays a well-regarded point tool while rivals bundle and automate around it, the modest funding stops looking like discipline.

How a challenger would attack it

Automate what the humans behind the curtain are doing. Kalepa’s most exploitable weakness is disclosed by its own employees: behind the Copilot interface sits a patchwork of third-party tools and a crowdsourced human review layer, run by a company described as too cost-conscious to buy enough software licenses. That is an operational cost structure a foundation-model-native entrant attacks directly — build the enrichment pipeline with current LLMs, no human reviewers, and undercut on both price and latency while marketing the contrast: “actually AI.” The second vector is capital and distribution asymmetry. Kalepa has raised $16M, hasn’t priced a round since September 2021, and sits at $5M estimated ARR; Sixfold ($50M+, Guidewire-backed) and Federato ($80M) can outspend it on every enterprise sales cycle, and Applied Systems’ back-to-back Planck and Cytora acquisitions mean the intake layer ships bundled with systems carriers already run. A challenger doesn’t need to beat Copilot’s accuracy everywhere — it needs to be good enough inside the core platform to make a standalone workbench an unjustifiable line item. The endgame attack is poaching the logos: Arch, Munich Re, and Berkley affiliates are precisely the accounts a well-funded rival targets with a free migration and an autonomous-agent roadmap Kalepa lacks the capital to counter-promise.

Same playbook, new buyer

Take the messy-submission engine to intermediaries and adjacent paper. Kalepa sells decision support to carriers, but the same capability — parsing broker emails, SOVs, and loss runs into structured, enriched risk pictures — is worth as much one step upstream: wholesale brokers and MGAs assembling submissions, who win on speed-to-market and currently do this assembly by hand. Kalepa already touches MGAs (Bowhead, Paragon, Beyond Risk) but as an underwriting tool; a broker-side product that packages cleaner, pre-enriched submissions would monetize the same models against a buyer the carrier-focused roadmap ignores. The second shift is line-of-business: Kalepa’s depth is US commercial P&C specialty; reinsurance treaty submissions, surety, and trade credit have the same unstructured-document chaos with no equivalent incumbent tooling and fewer competitors circling. The moat against Kalepa following is its own constraint set — ~58 people, no fresh capital in five years, and a land-and-expand motion committed to widening within existing carrier accounts. A company fighting to prove eight-figure ARR in its home market cannot open a second front, which leaves the adjacent buyers to whoever moves first.

Sources and further reading

• NYC-based insurance underwriting platform Kalepa raises $14M Series A led by Inspired Capital (TechCrunch, September 2021)
• Kalepa raises $14M Series A led by Inspired Capital (Kalepa, September 2021)
• The Story of Kalepa: Building the Future of Insurance Underwriting (Frontlines.io, 2023)
• From War Zones to Artificial Intelligence, Kalepa’s Daniel Hillman Knows the Power of Decision Making (Risk & Insurance, 2023)
• W. R. Berkley Corporation Expands Use of Kalepa’s Underwriting Platform (FF News, 2025)
• Kalepa’s Underwriting AI Platform Selected by Beyond Risk to Drive Scalable MGA Growth (FF News, 2026)
• Sixfold Raises US$30 Million Series B to Build the AI Underwriter (FF News, January 2026)
• Federato announces $80 million raised to bring RiskOps to insurance (PR Newswire, November 2024)
• Underwriting Software Market Size, Share, 2025-2030 Outlook (Mordor Intelligence, 2025)
• Kalepa Reviews — Pros & Cons of Working At Kalepa (Glassdoor, 2026)